Classify images with a pretrained ResNet-50 in ONNX Runtime, then adapt a pretrained network to your own categories.
You don't need to train a vision model from scratch to get good results. This course runs the ResNet-50 model from the hub with ONNX Runtime, explains the preprocessing that makes or breaks accuracy, and shows how transfer learning adapts a pretrained network to new categories.
Install with pip install onnxruntime numpy pillow.
4 lessons · 58 min
Pixels, convolutions and why pretrained networks transfer so well.
Resize, crop, scale and normalise — exactly as in training.
Run the model, turn scores into probabilities and read the top five.
Reuse a pretrained network for a new task with little data.
No ratings yet.
Sign in to leave a rating. Sign in
Use @handle to bring somebody in, and #tag to file it.
Nothing here yet. Ask a question, or say how you used it.
Sign in to join the discussion. Sign in
2 min read
2 min read
2 min read
2 min read
Transcribe and translate audio with OpenAI's open Whisper model, handle long recordings, and measure accuracy with word error rate.
Trend, seasonality, honest baselines and time-based testing, using NASA temperatures and bike rentals from the hub.
Where the model sits in the workflow, who checks it, and what happens to the exceptions.